DeepSeek raises $7B, Akamai jumps 20% on $1.8B AI deal

The AI infrastructure arms race entered a new phase on Monday, with three separate multi-billion-dollar commitments reshaping the competitive landscape. DeepSeek is raising over $7 billion to fund its revenue push, Akamai Technologies surged 20% after a leading U.S. frontier model provider committed $1.8 billion over seven years for its Cloud Infrastructure Services, and IREN Limited jumped 4% after announcing a deal with Nvidia to deploy up to 5 GW of AI infrastructure backed by a $2.1 billion Nvidia investment. The deals come as CoreWeave slid 13% after its Q2 revenue guidance of $2.45 billion to $2.6 billion fell short of the $2.69 billion consensus, signaling that even hyperscale winners face market skepticism. Rocket Lab rose 14% after its order backlog grew 20% quarter-over-quarter to a record $2.2 billion, while Upwork tumbled 21% on a restructuring that cuts 24% of its workforce. These moves underscore a market bifurcation: capital is flooding into infrastructure providers with locked-in demand, while companies reliant on discretionary AI spending face a reckoning. The $7 billion DeepSeek raise and the Akamai and IREN deals demonstrate that the AI buildout is accelerating, not slowing, and the winners are those with the most direct line to hyperscaler and frontier model budgets.

Where the $570M gap in CoreWeave's guidance matters

CoreWeave's 13% slide on Monday was the most telling signal of market discipline in the AI infrastructure sector. The company guided Q2 revenue to a range of $2.45 billion to $2.6 billion, missing the $2.69 billion consensus by as much as $240 million at the midpoint. The $570 million gap between the top of guidance and the consensus estimate represents a significant confidence shock for a stock that had been a bellwether for the AI cloud buildout. CoreWeave has been one of the most aggressive deployers of Nvidia GPUs, leasing data center capacity to AI startups and hyperscalers alike. The guidance miss shows that either demand is softening at the margin or that CoreWeave's capacity deployment is hitting operational bottlenecks. The company's business model depends on rapid GPU deployment and long-term contracts; any slowdown in either creates a compounding effect on revenue growth. The miss also raises questions about the sustainability of the AI cloud rental model, where providers like CoreWeave and Lambda Labs have raised billions to pre-purchase GPUs and then lease them out. If demand growth decelerates, the capital intensity of the model becomes a liability. CoreWeave's slide contrasted sharply with Akamai's 20% surge, highlighting that the market now differentiates sharply between infrastructure providers with diversified enterprise distribution and those concentrated on AI startup exposure. CoreWeave's Q2 guidance miss is the first major earnings disappointment from a pure-play AI infrastructure company, and it will force investors to re-examine the unit economics of the GPU-rental model. The company's reliance on short-term AI startup contracts, rather than long-term hyperscaler commitments, amplifies the risk when demand softens.

How Akamai's $1.8B deal rewrites its AI narrative

Akamai's 20% surge on Monday was driven by a single contract: a leading U.S. frontier model provider committed $1.8 billion over seven years for Akamai's Cloud Infrastructure Services. This deal transforms Akamai's position in the AI infrastructure market. Historically known as a content delivery network provider, Akamai has been investing in edge computing and cloud infrastructure, but it lacked the hyperscale narrative of AWS, Azure, or Google Cloud. The $1.8 billion commitment, spread over seven years, implies an annual run rate of roughly $257 million from a single customer. For a company that reported $3.98 billion in total revenue in 2025, this contract represents a meaningful new revenue stream. The seven-year duration also provides revenue visibility that most AI infrastructure companies lack. The identity of the frontier model provider is undisclosed, but the leading candidates include OpenAI, Anthropic, and Google DeepMind. The deal signals that frontier model companies are diversifying their cloud infrastructure away from the Big Three hyperscalers, seeking alternative providers with specialized capabilities. Akamai's edge network offers lower latency for inference workloads, which is critical for real-time AI applications. The contract also validates Akamai's strategy of building out compute capacity at the edge rather than in centralized data centers. For the broader market, the deal demonstrates that AI infrastructure demand is not limited to the hyperscalers; specialized providers with differentiated architectures can capture significant contracts. Akamai's stock move also reflects relief that the company has found a high-value AI use case for its network, after years of investor skepticism about its cloud ambitions.

IREN and Nvidia's 5 GW deal reshapes the power-constrained frontier

IREN Limited's 4% gain on Monday understates the significance of its deal with Nvidia: up to 5 GW of AI infrastructure deployment backed by a $2.1 billion Nvidia investment. This is one of the largest single commitments to AI data center capacity ever announced. To understand the scale, 5 GW is equivalent to the power output of five large nuclear reactors. The deal positions IREN as a major beneficiary of the power-constrained AI buildout. IREN, formerly known as Iris Energy, started as a Bitcoin miner and pivoted to AI data centers, leveraging its existing power infrastructure and site permits. Nvidia's $2.1 billion investment is not just a customer commitment; it is a strategic equity stake that aligns Nvidia's incentives with IREN's capacity deployment. Nvidia is effectively using its balance sheet to create demand for its own GPUs, ensuring that the 5 GW of capacity will be filled with Nvidia hardware. This deal also signals that the bottleneck for AI infrastructure is shifting from GPU supply to power availability. Data center power demand in the U.S. is projected to grow from 3% of total electricity consumption in 2022 to 8% by 2030, and sites with existing power permits and grid interconnections command a premium. IREN's Bitcoin mining background gave it access to low-cost power in locations like Texas and Canada, which are now prime AI data center sites. The deal also pressures competitors like CoreWeave, which has focused on GPU procurement rather than power infrastructure. IREN and Nvidia are creating a vertically integrated model where power, GPUs, and data center operations are bundled. This is the template for future AI infrastructure deals, as hyperscalers and GPU vendors race to secure power-constrained sites.

Why Rocket Lab's $2.2B backlog signals a space-AI convergence

Rocket Lab's 14% rise on Monday, driven by a record $2.2 billion order backlog that grew 20% quarter-over-quarter, points to a less obvious AI infrastructure play: space-based data processing and hypersonic testing. The company also won a $30 million contract for HASTE hypersonic test launches, which are critical for defense and commercial applications. Rocket Lab's backlog growth is notable because it includes contracts for its Neutron rocket, which is designed for medium-lift launches and satellite constellation deployment. The connection to AI is twofold. First, satellite constellations for Earth observation and communications generate massive amounts of data that require AI processing in orbit or at ground stations. Second, hypersonic testing generates terabytes of sensor data per flight, which requires AI-driven analysis. Rocket Lab's backlog growth shows that defense and commercial customers are accelerating their space-based AI infrastructure investments. The company's vertical integration, from rockets to satellite components, gives it a cost advantage over competitors like SpaceX and Relativity Space. The $30 million HASTE contract is particularly significant because hypersonic testing is a high-priority area for the U.S. Department of Defense, which is investing heavily in AI-powered missile defense and autonomous systems. Rocket Lab's backlog also includes contracts for NASA missions and commercial satellite operators, diversifying its revenue base. For the AI infrastructure theme, Rocket Lab represents the next frontier: space-based compute and data relay. As AI workloads move to the edge, including orbital platforms, companies that can launch and operate satellite constellations will capture value. Rocket Lab's 20% quarterly backlog growth outpaces most AI infrastructure companies, showing that the space-AI convergence is accelerating faster than the market realizes.

Tom Steyer's AI jobs guarantee and the California policy signal

California gubernatorial candidate Tom Steyer has proposed a jobs guarantee for workers displaced by AI, funded by a "token tax" on big tech companies for AI data processing. The proposal, reported by WIRED, would direct funds to the Golden State Sovereign Wealth Fund, which would finance jobs in housing, healthcare, and energy infrastructure. This is the most concrete policy proposal yet from a major political figure to address AI-driven labor displacement. Steyer's plan targets the core tension in the AI economy: the technology creates enormous value for companies and investors but threatens to displace millions of workers. The token tax would be levied on AI data processing, effectively taxing compute rather than profits. This is a novel approach that avoids the complexities of taxing AI output or robot labor. The proposal also connects AI policy to California's existing fiscal infrastructure, the sovereign wealth fund, which was established to manage budget surpluses from volatile tax revenues. Steyer's plan is a long-shot in a crowded gubernatorial field, but it signals that AI policy is becoming a central issue in the 2026 California governor's race. The state is home to most of the major AI companies, including OpenAI, Anthropic, and Google, and any policy changes there would have national implications. The token tax concept is particularly interesting because it targets the input of AI production (compute) rather than the output. This makes it easier to administer and harder to avoid than taxes on AI-generated content or services. For investors, the proposal adds regulatory risk to the AI infrastructure thesis. If California implements a token tax, it will increase the cost of AI compute in the state, potentially driving data center investment to other states or countries. Steyer's plan is still early-stage, but it represents the first serious attempt to create a fiscal framework for the AI economy. Whether California adopts a token tax or not, the political pressure it signals will force AI companies to accelerate their own workforce transition programs ahead of any binding regulation. Investors pricing AI infrastructure plays in California should now factor state-level policy risk into long-duration projects.